Wind speed

Load data

import pandas as pd
df_wind = pd.read_csv('../data/data_wind.csv')

/tmp/ipykernel_543748/1120575955.py:1: DeprecationWarning: 
Pyarrow will become a required dependency of pandas in the next major release of pandas (pandas 3.0),
(to allow more performant data types, such as the Arrow string type, and better interoperability with other libraries)
but was not found to be installed on your system.
If this would cause problems for you,
please provide us feedback at https://github.com/pandas-dev/pandas/issues/54466
        
  import pandas as pd

df_wind.head()

	HORA (UTC)	VENTO, DIRE��O HORARIA (gr) (� (gr))	VENTO, VELOCIDADE HORARIA (m/s)	UMIDADE REL. MAX. NA HORA ANT. (AUT) (%)	UMIDADE REL. MIN. NA HORA ANT. (AUT) (%)	TEMPERATURA M�XIMA NA HORA ANT. (AUT) (�C)	TEMPERATURA M�NIMA NA HORA ANT. (AUT) (�C)	UMIDADE RELATIVA DO AR, HORARIA (%)	PRESSAO ATMOSFERICA AO NIVEL DA ESTACAO, HORARIA (mB)	PRECIPITA��O TOTAL, HOR�RIO (mm)	VENTO, RAJADA MAXIMA (m/s)	PRESS�O ATMOSFERICA MAX.NA HORA ANT. (AUT) (mB)	PRESS�O ATMOSFERICA MIN. NA HORA ANT. (AUT) (mB)
0	12:00	0.809017	1.8	69.0	60.0	22.6	20.7	61.0	888.2	0.0	3.8	888.2	887.7
1	13:00	0.965926	2.7	62.0	55.0	24.2	22.5	55.0	888.4	0.0	4.7	888.4	888.2
2	14:00	0.891007	2.0	56.0	50.0	25.5	24.3	51.0	888.1	0.0	4.9	888.4	888.1
3	15:00	0.848048	2.5	52.0	44.0	27.4	25.0	44.0	887.4	0.0	5.8	888.1	887.4
4	16:00	0.224951	2.4	50.0	43.0	27.1	25.5	46.0	886.5	0.0	5.8	887.4	886.5

df_wind.info()

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 87693 entries, 0 to 87692
Data columns (total 13 columns):
 #   Column                                                 Non-Null Count  Dtype  
---  ------                                                 --------------  -----  
 0   HORA (UTC)                                             87693 non-null  object 
 1   VENTO, DIRE��O HORARIA (gr) (� (gr))             87693 non-null  float64
 2   VENTO, VELOCIDADE HORARIA (m/s)                        87693 non-null  float64
 3   UMIDADE REL. MAX. NA HORA ANT. (AUT) (%)               87693 non-null  float64
 4   UMIDADE REL. MIN. NA HORA ANT. (AUT) (%)               87693 non-null  float64
 5   TEMPERATURA M�XIMA NA HORA ANT. (AUT) (�C)         87693 non-null  float64
 6   TEMPERATURA M�NIMA NA HORA ANT. (AUT) (�C)         87693 non-null  float64
 7   UMIDADE RELATIVA DO AR, HORARIA (%)                    87693 non-null  float64
 8   PRESSAO ATMOSFERICA AO NIVEL DA ESTACAO, HORARIA (mB)  87693 non-null  float64
 9   PRECIPITA��O TOTAL, HOR�RIO (mm)                 87693 non-null  float64
 10  VENTO, RAJADA MAXIMA (m/s)                             87693 non-null  float64
 11  PRESS�O ATMOSFERICA MAX.NA HORA ANT. (AUT) (mB)      87693 non-null  float64
 12  PRESS�O ATMOSFERICA MIN. NA HORA ANT. (AUT) (mB)     87693 non-null  float64
dtypes: float64(12), object(1)
memory usage: 8.7+ MB

df_wind.isnull().sum().sum()

0

We don’t have missing values.

df_wind.columns

Index(['HORA (UTC)', 'VENTO, DIRE��O HORARIA (gr) (� (gr))',
       'VENTO, VELOCIDADE HORARIA (m/s)',
       'UMIDADE REL. MAX. NA HORA ANT. (AUT) (%)',
       'UMIDADE REL. MIN. NA HORA ANT. (AUT) (%)',
       'TEMPERATURA M�XIMA NA HORA ANT. (AUT) (�C)',
       'TEMPERATURA M�NIMA NA HORA ANT. (AUT) (�C)',
       'UMIDADE RELATIVA DO AR, HORARIA (%)',
       'PRESSAO ATMOSFERICA AO NIVEL DA ESTACAO, HORARIA (mB)',
       'PRECIPITA��O TOTAL, HOR�RIO (mm)', 'VENTO, RAJADA MAXIMA (m/s)',
       'PRESS�O ATMOSFERICA MAX.NA HORA ANT. (AUT) (mB)',
       'PRESS�O ATMOSFERICA MIN. NA HORA ANT. (AUT) (mB)'],
      dtype='object')

Preprocessing

df_wind_preprocessed = df_wind.copy()

We don’t have significative column names so we can rename them.

new_columns = [
    'Hora',
    'Dirección del viento horaria (gr)',
    'Velocidad horario del viento (m/s)',
    'Humedad rel. máx. hora anterior (AUT) (%)',
    'Humedad rel. mín. hora anterior (AUT) (%)',
    'Temperatura máx. hora anterior (AUT) (℃)',
    'Temperatura mín. hora anterior (AUT) (℃)',
    'Humedad relativa horaria (%)',
    'Presión atmosférica a nivel de estación, horaria (mB)',
    'Precipitación total por hora (mm)',
    'Máxima ráfaga de viento (m/s)',
    'Presión atmosférica máx. hora anterior (AUT) (mB)',
    'Presión atmosférica mín. hora anterior (AUT) (mB)'
]

df_wind_preprocessed.rename(columns=dict(zip(df_wind.columns, new_columns)), inplace=True)
df_wind_preprocessed.columns

Index(['Hora', 'Dirección del viento horaria (gr)',
       'Velocidad horario del viento (m/s)',
       'Humedad rel. máx. hora anterior (AUT) (%)',
       'Humedad rel. mín. hora anterior (AUT) (%)',
       'Temperatura máx. hora anterior (AUT) (℃)',
       'Temperatura mín. hora anterior (AUT) (℃)',
       'Humedad relativa horaria (%)',
       'Presión atmosférica a nivel de estación, horaria (mB)',
       'Precipitación total por hora (mm)', 'Máxima ráfaga de viento (m/s)',
       'Presión atmosférica máx. hora anterior (AUT) (mB)',
       'Presión atmosférica mín. hora anterior (AUT) (mB)'],
      dtype='object')

print(df_wind_preprocessed['Hora'][0])
type(df_wind_preprocessed['Hora'][0])

12:00

str

Some values are missing, sometimes we go from the hour 17 to 19 for exemple. We can add this missing hours to the dataset.

total_added = 0

for _ in range(23):
    hours = df_wind_preprocessed['Hora']
    idx = 0
    nb_added = 0

    while idx < len(hours) - 1:
        cur_hour = int(hours[idx].split(':')[0])
        next_hour = int(hours[idx + 1].split(':')[0])
        if next_hour != (cur_hour + 1) % 24:
            # print(cur_hour, next_hour, f"=> add {str((cur_hour + 1) % 24)+':00'}")
            if (cur_hour + 1) % 24 < 10:
                new_str = "0"+str((cur_hour + 1) % 24)+":00"
            else:
                new_str = str((cur_hour + 1) % 24)+":00"
            new_row = pd.DataFrame({'Hora': [new_str]})
            df_wind_preprocessed = pd.concat([df_wind_preprocessed.iloc[:idx + nb_added + 1], new_row, df_wind_preprocessed.iloc[idx + nb_added + 1:]], ignore_index=True)
            nb_added += 1
        idx += 1
    print(f"Added {nb_added} rows")
    total_added += nb_added

print(f"Added {total_added} en total")

Added 866 rows

Added 332 rows

Added 241 rows

Added 182 rows

Added 141 rows

Added 122 rows

Added 106 rows

Added 92 rows

Added 79 rows

Added 67 rows

Added 57 rows

Added 52 rows

Added 44 rows

Added 41 rows

Added 37 rows

Added 31 rows

Added 27 rows

Added 23 rows

Added 16 rows

Added 11 rows

Added 8 rows

Added 5 rows

Added 3 rows
Added 2583 en total

The hour is given as a string. Considered we have data for almost ten years, we can change the format of the hour, starting with a given year, for instance the year 2000.

import datetime as dt

start_date = dt.datetime(2000, 1, 1, int(df_wind_preprocessed['Hora'][0][:2]))

df_wind_preprocessed['Fecha'] = [start_date + dt.timedelta(hours=i) for i in range(len(df_wind_preprocessed['Hora']))]

The data are not in the units of the international system. We can change that.

df_wind_preprocessed['Hora'] = df_wind_preprocessed['Hora'].apply(lambda x: int(x.split(':')[0]))

df_wind_preprocessed['Humedad rel. máx. hora anterior (AUT)'] = df_wind_preprocessed['Humedad rel. máx. hora anterior (AUT) (%)'].apply(lambda x: x/100)
df_wind_preprocessed.drop(columns=['Humedad rel. máx. hora anterior (AUT) (%)'], inplace=True)

df_wind_preprocessed['Humedad rel. mín. hora anterior (AUT)'] = df_wind_preprocessed['Humedad rel. mín. hora anterior (AUT) (%)'].apply(lambda x: x/100)
df_wind_preprocessed.drop(columns=['Humedad rel. mín. hora anterior (AUT) (%)'], inplace=True)

df_wind_preprocessed['Temperatura máx. hora anterior (AUT) (K)'] = df_wind_preprocessed['Temperatura máx. hora anterior (AUT) (℃)'].apply(lambda x: x+273.15)
df_wind_preprocessed.drop(columns=['Temperatura máx. hora anterior (AUT) (℃)'], inplace=True)

df_wind_preprocessed['Temperatura mín. hora anterior (AUT) (K)'] = df_wind_preprocessed['Temperatura mín. hora anterior (AUT) (℃)'].apply(lambda x: x+273.15)
df_wind_preprocessed.drop(columns=['Temperatura mín. hora anterior (AUT) (℃)'], inplace=True)

df_wind_preprocessed['Humedad relativa horaria'] = df_wind_preprocessed['Humedad relativa horaria (%)'].apply(lambda x: x/100)
df_wind_preprocessed.drop(columns=['Humedad relativa horaria (%)'], inplace=True)

df_wind_preprocessed['Presión atmosférica a nivel de estación, horaria (Pa)'] = df_wind_preprocessed['Presión atmosférica a nivel de estación, horaria (mB)'].apply(lambda x: x*100)
df_wind_preprocessed.drop(columns=['Presión atmosférica a nivel de estación, horaria (mB)'], inplace=True)

df_wind_preprocessed['Precipitación total por hora (m)'] = df_wind_preprocessed['Precipitación total por hora (mm)'].apply(lambda x: x/1000)
df_wind_preprocessed.drop(columns=['Precipitación total por hora (mm)'], inplace=True)

df_wind_preprocessed['Presión atmosférica máx. hora anterior (AUT) (Pa)'] = df_wind_preprocessed['Presión atmosférica máx. hora anterior (AUT) (mB)'].apply(lambda x: x*100)
df_wind_preprocessed.drop(columns=['Presión atmosférica máx. hora anterior (AUT) (mB)'], inplace=True)

df_wind_preprocessed['Presión atmosférica mín. hora anterior (AUT) (Pa)'] = df_wind_preprocessed['Presión atmosférica mín. hora anterior (AUT) (mB)'].apply(lambda x: x*100)
df_wind_preprocessed.drop(columns=['Presión atmosférica mín. hora anterior (AUT) (mB)'], inplace=True)

df_wind_preprocessed.columns

Index(['Hora', 'Dirección del viento horaria (gr)',
       'Velocidad horario del viento (m/s)', 'Máxima ráfaga de viento (m/s)',
       'Fecha', 'Humedad rel. máx. hora anterior (AUT)',
       'Humedad rel. mín. hora anterior (AUT)',
       'Temperatura máx. hora anterior (AUT) (K)',
       'Temperatura mín. hora anterior (AUT) (K)', 'Humedad relativa horaria',
       'Presión atmosférica a nivel de estación, horaria (Pa)',
       'Precipitación total por hora (m)',
       'Presión atmosférica máx. hora anterior (AUT) (Pa)',
       'Presión atmosférica mín. hora anterior (AUT) (Pa)'],
      dtype='object')

df_wind_preprocessed.head()

	Hora	Dirección del viento horaria (gr)	Velocidad horario del viento (m/s)	Máxima ráfaga de viento (m/s)	Fecha	Humedad rel. máx. hora anterior (AUT)	Humedad rel. mín. hora anterior (AUT)	Temperatura máx. hora anterior (AUT) (K)	Temperatura mín. hora anterior (AUT) (K)	Humedad relativa horaria	Presión atmosférica a nivel de estación, horaria (Pa)	Precipitación total por hora (m)	Presión atmosférica máx. hora anterior (AUT) (Pa)	Presión atmosférica mín. hora anterior (AUT) (Pa)
0	12	0.809017	1.8	3.8	2000-01-01 12:00:00	0.69	0.60	295.75	293.85	0.61	88820.0	0.0	88820.0	88770.0
1	13	0.965926	2.7	4.7	2000-01-01 13:00:00	0.62	0.55	297.35	295.65	0.55	88840.0	0.0	88840.0	88820.0
2	14	0.891007	2.0	4.9	2000-01-01 14:00:00	0.56	0.50	298.65	297.45	0.51	88810.0	0.0	88840.0	88810.0
3	15	0.848048	2.5	5.8	2000-01-01 15:00:00	0.52	0.44	300.55	298.15	0.44	88740.0	0.0	88810.0	88740.0
4	16	0.224951	2.4	5.8	2000-01-01 16:00:00	0.50	0.43	300.25	298.65	0.46	88650.0	0.0	88740.0	88650.0

df_wind_preprocessed.to_csv('../data/data_wind_preprocessed.csv', index=False)

Load data preprocessed

import pandas as pd

data_wind = pd.read_csv("../data/data_wind_preprocessed.csv")

EDA

data_eda = data_wind.drop(columns=['Hora', 'Fecha'])
data_eda.shape

(90276, 12)

Missing values

missing_data_count = data_wind.isnull().sum()
missing_data_percentage = round((missing_data_count / len(data_wind)) * 100, 1)

missing_data_stats = pd.DataFrame({
    'Missing data (count)': missing_data_count,
    'Missing data (%)': missing_data_percentage
})

missing_data_stats

	Missing data (count)	Missing data (%)
Hora	0	0.0
Dirección del viento horaria (gr)	2583	2.9
Velocidad horario del viento (m/s)	2583	2.9
Máxima ráfaga de viento (m/s)	2583	2.9
Fecha	0	0.0
Humedad rel. máx. hora anterior (AUT)	2583	2.9
Humedad rel. mín. hora anterior (AUT)	2583	2.9
Temperatura máx. hora anterior (AUT) (K)	2583	2.9
Temperatura mín. hora anterior (AUT) (K)	2583	2.9
Humedad relativa horaria	2583	2.9
Presión atmosférica a nivel de estación, horaria (Pa)	2583	2.9
Precipitación total por hora (m)	2583	2.9
Presión atmosférica máx. hora anterior (AUT) (Pa)	2583	2.9
Presión atmosférica mín. hora anterior (AUT) (Pa)	2583	2.9

Statistics

data_wind.describe()

	Hora	Dirección del viento horaria (gr)	Velocidad horario del viento (m/s)	Máxima ráfaga de viento (m/s)	Humedad rel. máx. hora anterior (AUT)	Humedad rel. mín. hora anterior (AUT)	Temperatura máx. hora anterior (AUT) (K)	Temperatura mín. hora anterior (AUT) (K)	Humedad relativa horaria	Presión atmosférica a nivel de estación, horaria (Pa)	Precipitación total por hora (m)	Presión atmosférica máx. hora anterior (AUT) (Pa)	Presión atmosférica mín. hora anterior (AUT) (Pa)
count	90276.000000	87693.000000	87693.000000	87693.000000	87693.000000	87693.000000	87693.000000	87693.000000	87693.000000	87693.000000	87693.000000	87693.000000	87693.000000
mean	11.500798	0.405810	2.466192	5.161076	0.690585	0.631762	295.071264	293.834570	0.661467	88725.192547	0.000161	88758.072366	88689.109279
std	6.922225	0.686247	1.313968	2.311157	0.196402	0.201663	3.721386	3.513744	0.199923	401.240375	0.001308	364.675005	356.453928
min	0.000000	-1.000000	0.000000	0.000000	0.120000	0.100000	282.350000	281.550000	0.100000	86340.000000	0.000000	86530.000000	86280.000000
25%	6.000000	-0.156434	1.500000	3.400000	0.540000	0.480000	292.350000	291.550000	0.510000	88530.000000	0.000000	88560.000000	88500.000000
50%	12.000000	0.788011	2.400000	5.000000	0.720000	0.640000	294.550000	293.350000	0.680000	88720.000000	0.000000	88750.000000	88690.000000
75%	18.000000	0.970296	3.400000	6.800000	0.870000	0.800000	297.850000	296.250000	0.840000	88910.000000	0.000000	88930.000000	88880.000000
max	23.000000	1.000000	10.000000	24.300000	1.000000	0.980000	308.450000	307.550000	0.990000	102350.000000	0.070800	91310.000000	91090.000000

Plot time sery

import plotly.graph_objects as go

for col in data_eda.columns:
    fig = go.Figure()

    fig.add_trace(go.Scatter(x=data_wind['Fecha'], y=data_wind[col], mode='lines', name=col))

    fig.update_layout(title=f'{col} over time',
                  xaxis_title='Date', 
                  yaxis_title=col,
                  margin=dict(l=20, r=20, t=40, b=20))
    fig.show()

• We can see that the wind direction change a lot and very fast.

• We can see cycles during the year in the humidity (min, max and relative) and cycles during the day for temperatures but not cycles during the year so we can assume a special climate.

• We can see outliers for the atmosferical pressure. Because of the type of validation, we are going to keep them because it doesn’t influence all validations, only a few ones.

• We can notice a very high variability for some values so we can imagine some imprecisions of measure for some periods.

• We can see cycles for the atmosferical pressure during the days and during the years.

Random week

from matplotlib import pyplot as plt
from random import randint, seed
seed(42)

fig, axs = plt.subplots(12, 1, figsize=(20, 60))
nb_values = 24*7

rdm_idx = randint(0, data_wind.shape[0] - nb_values)

for i, column in enumerate(data_eda.columns):
    axs[i].plot(data_wind['Fecha'][rdm_idx:rdm_idx+nb_values], data_eda[column][rdm_idx:rdm_idx+nb_values])
    axs[i].set_title(column)

plt.tight_layout()
plt.show()

At the scale of a week, we can confirm a lot of stationarity.

Boxplot

import seaborn as sns
import matplotlib.pyplot as plt

fig, axs = plt.subplots(4, 3, figsize=(20, 20))

for i, column in enumerate(data_eda.columns):
    sns.boxplot(data=data_eda[column], ax=axs[i//3, i%3])
    axs[i//3, i%3].set_title(column)

plt.tight_layout()
plt.show()

We can see a lot of outliers.
For the total precipitation, the high majority of values are concentrated around 0 which means we have a lot of hours without rain.

Let’s see the correlation of the variables with the output.

Scatterplot and regplot

import seaborn as sns
import matplotlib.pyplot as plt

fig, axs = plt.subplots(4, 3, figsize=(20, 20))

for i, column in enumerate(data_eda.columns):
    sns.regplot(data=data_eda, x='Velocidad horario del viento (m/s)', y=column, ax=axs[i//3, i%3])
    axs[i//3, i%3].set_title(column)

plt.tight_layout()
plt.show()

---------------------------------------------------------------------------
KeyboardInterrupt                         Traceback (most recent call last)
Cell In[21], line 7
      4 fig, axs = plt.subplots(4, 3, figsize=(20, 20))
      6 for i, column in enumerate(data_eda.columns):
----> 7     sns.regplot(data=data_eda, x='Velocidad horario del viento (m/s)', y=column, ax=axs[i//3, i%3])
      8     axs[i//3, i%3].set_title(column)
     10 plt.tight_layout()

File ~/miniconda3/envs/ml_venv/lib/python3.9/site-packages/seaborn/regression.py:775, in regplot(data, x, y, x_estimator, x_bins, x_ci, scatter, fit_reg, ci, n_boot, units, seed, order, logistic, lowess, robust, logx, x_partial, y_partial, truncate, dropna, x_jitter, y_jitter, label, color, marker, scatter_kws, line_kws, ax)
    773 scatter_kws["marker"] = marker
    774 line_kws = {} if line_kws is None else copy.copy(line_kws)
--> 775 plotter.plot(ax, scatter_kws, line_kws)
    776 return ax

File ~/miniconda3/envs/ml_venv/lib/python3.9/site-packages/seaborn/regression.py:384, in _RegressionPlotter.plot(self, ax, scatter_kws, line_kws)
    381     self.scatterplot(ax, scatter_kws)
    383 if self.fit_reg:
--> 384     self.lineplot(ax, line_kws)
    386 # Label the axes
    387 if hasattr(self.x, "name"):

File ~/miniconda3/envs/ml_venv/lib/python3.9/site-packages/seaborn/regression.py:429, in _RegressionPlotter.lineplot(self, ax, kws)
    427 """Draw the model."""
    428 # Fit the regression model
--> 429 grid, yhat, err_bands = self.fit_regression(ax)
    430 edges = grid[0], grid[-1]
    432 # Get set default aesthetics

File ~/miniconda3/envs/ml_venv/lib/python3.9/site-packages/seaborn/regression.py:229, in _RegressionPlotter.fit_regression(self, ax, x_range, grid)
    227     yhat, yhat_boots = self.fit_logx(grid)
    228 else:
--> 229     yhat, yhat_boots = self.fit_fast(grid)
    231 # Compute the confidence interval at each grid point
    232 if ci is None:

File ~/miniconda3/envs/ml_venv/lib/python3.9/site-packages/seaborn/regression.py:250, in _RegressionPlotter.fit_fast(self, grid)
    247 if self.ci is None:
    248     return yhat, None
--> 250 beta_boots = algo.bootstrap(X, y,
    251                             func=reg_func,
    252                             n_boot=self.n_boot,
    253                             units=self.units,
    254                             seed=self.seed).T
    255 yhat_boots = grid.dot(beta_boots).T
    256 return yhat, yhat_boots

File ~/miniconda3/envs/ml_venv/lib/python3.9/site-packages/seaborn/algorithms.py:100, in bootstrap(*args, **kwargs)
     98     resampler = integers(0, n, n, dtype=np.intp)  # intp is indexing dtype
     99     sample = [a.take(resampler, axis=0) for a in args]
--> 100     boot_dist.append(f(*sample, **func_kwargs))
    101 return np.array(boot_dist)

File ~/miniconda3/envs/ml_venv/lib/python3.9/site-packages/seaborn/regression.py:242, in _RegressionPlotter.fit_fast.<locals>.reg_func(_x, _y)
    241 def reg_func(_x, _y):
--> 242     return np.linalg.pinv(_x).dot(_y)

File ~/miniconda3/envs/ml_venv/lib/python3.9/site-packages/numpy/linalg/linalg.py:2030, in pinv(a, rcond, hermitian)
   2027 s = divide(1, s, where=large, out=s)
   2028 s[~large] = 0
-> 2030 res = matmul(transpose(vt), multiply(s[..., newaxis], transpose(u)))
   2031 return wrap(res)

KeyboardInterrupt: 

In general there is not a clear linear correlation between the values and the output except for the pressure and the wind gusts.

We are going to use an linear imputation because a value is high correlated to the last and the next value.

Imputation of data

import warnings
warnings.filterwarnings('ignore')

data_wind_imputed = data_wind.interpolate(method='linear', axis=0)

data_wind_imputed.to_csv('../data/data_wind_imputed.csv', index=False)

Before imputation

data_wind[317:320]

	Hora	Dirección del viento horaria (gr)	Velocidad horario del viento (m/s)	Máxima ráfaga de viento (m/s)	Fecha	Humedad rel. máx. hora anterior (AUT)	Humedad rel. mín. hora anterior (AUT)	Temperatura máx. hora anterior (AUT) (K)	Temperatura mín. hora anterior (AUT) (K)	Humedad relativa horaria	Presión atmosférica a nivel de estación, horaria (Pa)	Precipitación total por hora (m)	Presión atmosférica máx. hora anterior (AUT) (Pa)	Presión atmosférica mín. hora anterior (AUT) (Pa)
317	17	0.838671	2.9	6.4	2000-01-14 17:00:00	0.54	0.48	298.85	297.35	0.49	89000.0	0.0	89050.0	89000.0
318	18	NaN	NaN	NaN	2000-01-14 18:00:00	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN
319	19	0.766044	2.7	6.1	2000-01-14 19:00:00	0.51	0.45	299.15	298.05	0.47	88950.0	0.0	88960.0	88940.0

After imputation

data_wind_imputed[317:320]

	Hora	Dirección del viento horaria (gr)	Velocidad horario del viento (m/s)	Máxima ráfaga de viento (m/s)	Fecha	Humedad rel. máx. hora anterior (AUT)	Humedad rel. mín. hora anterior (AUT)	Temperatura máx. hora anterior (AUT) (K)	Temperatura mín. hora anterior (AUT) (K)	Humedad relativa horaria	Presión atmosférica a nivel de estación, horaria (Pa)	Precipitación total por hora (m)	Presión atmosférica máx. hora anterior (AUT) (Pa)	Presión atmosférica mín. hora anterior (AUT) (Pa)
317	17	0.838671	2.9	6.40	2000-01-14 17:00:00	0.540	0.480	298.85	297.35	0.49	89000.0	0.0	89050.0	89000.0
318	18	0.802358	2.8	6.25	2000-01-14 18:00:00	0.525	0.465	299.00	297.70	0.48	88975.0	0.0	89005.0	88970.0
319	19	0.766044	2.7	6.10	2000-01-14 19:00:00	0.510	0.450	299.15	298.05	0.47	88950.0	0.0	88960.0	88940.0